Once Optimus production reaches a scale of a million units per year, the marginal production cost will be roughly $20,000-25,000 per robot.
“I think at volume at a million units a year, the the production cost is probably on the order of $20,000, maybe 25, something like that.”
View on YouTube
Explanation
Optimus has not yet reached a production scale of a million units per year, so this cost claim cannot yet be verified; current per-unit manufacturing cost estimates are substantially higher, around $50,000-100,000.
Tesla's upcoming AI5 inference chip will deliver up to a 40x improvement over the current AI4 chip on key metrics, including an 8x raw compute increase compounded with further silicon-level optimizations.
“we're finaliz finalizing the design of AI5 which will be an immense jump from AI4. Um by some metrics the improvement in AI5 will be 40 times better than AI4.”
View on YouTube
Explanation
Tesla's AI5 chip, taped out in April 2026, was described by Musk as delivering roughly 8x compute, 9x memory, and 5x bandwidth improvements over AI4, broadly consistent with the up-to-40x combined improvement claim.
Tesla's FSD software version 14, releasing over the next few months as the biggest software upgrade since version 12, will make existing AI4-equipped cars achieve self-driving safety at least 2-3x (potentially up to 10x) that of a human driver, and cars will feel sentient by the end of the year.
“I I I am confident that the current uh chips uh AI AI4 chips that are in the cars will uh achieve self-driving safety that is at least two to three times that of of human and and maybe even 10x... the software that uh will be released for that is is coming out over the next few months. So version 14 will be the biggest uh upgrade in Tesla software since version 12... your car is going to feel like it is sentient by the end of the year.”
View on YouTube
Explanation
FSD version 14 did ship as a major upgrade, but Musk's specific safety-multiplier and 'sentient by end of year' claims are subjective and not independently verified.
Phones with chipsets modified to support the newly acquired Echostar spectrum for direct Starlink-to-phone connectivity will start shipping in approximately 2 years (around late 2027).
“the chipset has to be modified to add these frequencies um and that probably is a 2-year time frame. So the phones that um are able to use the spectrum that was acquired probably start shipping in around 2 years.”
View on YouTube
Explanation
This prediction concerns phones with modified chipsets shipping around late 2027, a timeframe not yet reached.
Barring major setbacks, SpaceX will demonstrate full Starship reusability next year, catching both the booster and the ship, while delivering over 100 tons to a useful orbit.
“unless we have unless we have some very major setbacks, uh, SpaceX will demonstrate uh, full reusability next year uh, catching both the booster and the ship um, and being able to deliver over 100 tons to a useful orbit.”
View on YouTube
Explanation
As of mid-2026, SpaceX had achieved booster catches but had not yet completed a full upper-stage tower catch (targeted for an upcoming flight) or demonstrated complete full reusability, and the predicted year is not yet over.
AI could be smarter than any single human at any task as soon as next year (2026).
“I I I I think that we might have AI smarter than any single human at anything as soon as next year.”
View on YouTube
Explanation
Whether AI surpassed any single human at any task by 2026 is a subjective and contested claim without clear consensus verification.
Within about five years, by roughly 2030, AI will likely be smarter than the combined intelligence of all humans.
“and and and then and then probably within five like say 2030 probably AI is smarter than the sum of all humans.”
View on YouTube
Explanation
This prediction concerns AI surpassing combined human intelligence by around 2030, a timeframe not yet reached.
If tonnage delivered to Mars increases exponentially with each biennial Mars transfer window, Mars could become self-sustaining (able to survive without resupply from Earth) in roughly 25 years, over 10-15 transfer windows.
“if you're um seeing exponential increases in the tonnage to Mars with each Mars transfer window, then it should be possible to make Mars self-sustaining um in in about call it roughly 25 years.”
View on YouTube
Explanation
This prediction concerns Mars becoming self-sustaining roughly 25 years out, a timeframe far in the future.
At the current pace of regulatory delay, the California high-speed rail project might not finish until sometime next century.
“at the current Trend you know California High Street rail might finish sometime next Century maybe probably not we're just we'll have teleportation by that time”
View on YouTube
Explanation
California High-Speed Rail construction has continued to face major delays and cost overruns through 2025-2026, with full completion still not achieved and current estimates pushing full system completion well past originally planned dates, broadly consistent with the pessimistic framing though 'next century' was hyperbole.
If Trump wins and Musk leads a government-efficiency effort, government spending/size could be reduced by a low single-digit percentage each year over four years.
“I was saying low low digits every year for four years would be palatable”
View on YouTube
Explanation
DOGE (the Trump-era Department of Government Efficiency effort Musk led in 2025) pursued aggressive spending and workforce cuts, but it fell well short of achieving sustained multi-year low-single-digit annual spending reductions, and Musk himself departed DOGE and the administration acrimoniously by mid-2025 before such a multi-year plan could be realized.
The upcoming Polaris Dawn mission (imminent at time of interview) will be the first private/commercial spacewalk and will reach the highest altitude since Apollo, the furthest from Earth anyone has traveled since then.
“this will be the first time uh the first private first first commercial space walk and and it be at at the highest altitude uh since Apollo so it's the furthest from Earth that anyone's gone”
View on YouTube
Explanation
Polaris Dawn launched in September 2024 and did achieve the first private/commercial spacewalk, reaching the highest Earth orbit altitude since the Apollo program.
SpaceX's Starship will launch to Mars (uncrewed) within two years, with an aspirational crewed launch to Mars about four years out, in the next available launch window.
“announced just the other day Starship going to Mars in two years and by the way huh yeah yeah yeah yeah yeah and then four years for crude uh aspirational launch in the next window”
View on YouTube
Explanation
SpaceX's Starship did not launch an uncrewed mission to Mars within two years of this August 2024 interview; the 2026 Mars transfer window passed without a Starship Mars launch, and Musk has repeatedly pushed back Starship's Mars timeline.
Musk estimates roughly an 80% probability that AI leads to a good future of abundance, versus about a 20% probability of an existential/annihilation-level bad outcome.
“that's probably 80% likely so look on the right side only 20% 20% probability of annihilation nothing”
View on YouTube
Explanation
This is a long-range probabilistic forecast about AI outcomes (utopia versus existential catastrophe) that cannot be resolved with current information.
Tesla's Dojo 2 training computer will be in volume production towards the end of next year (2025), roughly comparable to an Nvidia B200-type training system.
“we should have Dojo 2 in volume towards the end of next”
View on YouTube
Explanation
Tesla did not reach volume production of Dojo 2 by the end of 2025; instead, Tesla dissolved its Dojo supercomputer program in August 2025, shifting to Nvidia, AMD, and Samsung chips, before Musk announced a surprising revival of a Dojo 3 effort in January 2026.
Tesla will reach its third major iteration of the Dojo training computer (Dojo 3) by roughly late 2026.
“we'll only have the second major iteration next year um the third iteration I don't know maybe late you know 26 or something like that”
View on YouTube
Explanation
Tesla did not steadily progress to a 'Dojo 3' iteration by late 2026 on the originally described timeline; the entire Dojo program was shut down in August 2025 before being unexpectedly revived focused on space-based compute in January 2026, a very different trajectory than predicted.
At scale production (three major design iterations, million-plus units/year), Tesla's Optimus humanoid robot will cost around $20,000, less than a small car.
“so if a small car you know costs $25,000 you know it's it's probably like I don't know $20,000 for for an Optimus for a humanoid robot”
View on YouTube
Explanation
Optimus has not yet reached the scale production (million-plus units/year, three design iterations) needed to test this cost prediction, as of 2026.
It will take on the order of five to six years for Tesla's Optimus humanoid robot program to reach production scale of about a million units per year.
“I I would say m major iterations are less than two years so okay um it's probably on the order of five five years yeah uh maybe six to get to a million units a year”
View on YouTube
Explanation
Tesla's Optimus program has not yet reached anywhere close to a million units per year of production as of 2026, so this five-to-six-year timeline (through roughly 2029-2030) remains unresolved.
Within about 30 years, the number of humanoid robots on Earth will vastly exceed the number of humans, with a ratio of at least 2:1 and possibly 3:1.
“I think the ratio of humanoid robots to humans is going to be at least 2 to one maybe 3 to one because everybody every everybody will want one”
View on YouTube
Explanation
This is a roughly 30-year projection about humanoid robot population ratios that cannot be evaluated this early.
The rate of notable mistakes or "gaffes" in Elon's management of Twitter will decrease over time relative to the initial weeks after his acquisition in late 2022.
“I think generally like my error rate in sort of being the chief twit, uh, will be less over time. ... I think we'll have fewer, um, fewer gaffes in the future.”
View on YouTube
Measured hate-speech impressions on Twitter, which Elon claims are already down by about one-third as of December 2022, will decline further (i.e., fall below that level) over time following his acquisition.
“You're doing a lot more to take down hate speech than the company previously was doing.
Yeah, absolutely. Hate speech impressions are down by a third and will get even lower.”
View on YouTube
By mid to late January 2023, Twitter will roll out a new feature set that includes in-product information explaining why an account is suspended or subject to visibility filtering (shadow banning).
“something I've committed to and we'll, uh, I think probably be able to roll that out in January. ... the next sort of feature set will probably roll out mid to late January. Um, and hopefully in that we'll, we can include, um, information about why an account is suspended or has, uh, what is called within Tesla Twitter, uh, visibility filtering uh aka shadow banning.”
View on YouTube
The global (or at least U.S.) economy will enter a recession in 2023, with severity at least light-to-moderate and roughly comparable in magnitude to the 2009 recession.
“Well, I think it's more like it does seem like we're headed into a recession here, um, in 2023. The magnitude of that recession is debatable, but I think it's at least a light to moderate recession potentially. It's on the order of 2009.”
View on YouTube
Economic weakness or recessionary conditions will persist for roughly 1–2 years from late 2022, with clear recovery beginning around Q2 2024.
“my best guess is that, you know, we have sort of stormy times for a year or two, a year and a half. And then things start to dawn breaks roughly in Q2 24.”
View on YouTube